Extraction of the Fundamental Component in Speech with. a Nonlinear Circuit Model for the Cochlea
John L. Stewart · The Journal of the Acoustical Society of America · 1974
Distributed nonlinearities in a circuit model for basilar membrane vibrations demodulate the envelope component in speech in a cumulative way. Induced components propagate to the normal place of localization where they may be extracted. Waveforms and recordings for this component are demonstrated with well-known sentences under various conditions of speech pre-filter bandwidths and imposed noise. The extracted fundamental is comparatively reliable and tolerant of speech bandwidth variations. Specific demonstrated bandwidths are 30–5000 Hz, 300–3000 Hz, and 600–3000 Hz. Raising the lower frequency cutoff has little effect on voicing extraction except for a gradual weakening of /u/, /i/, and the nasals. Extending upper cutoff above 3000 Hz has little value. Experiments using the extracted voicing component as a tactile deaf aid are described, with plosive cues also present.